Comparison of Passenger Rail Energy Consumption with Competing Modes (2015)

Below is the uncorrected machine-read text of this chapter, intended to provide our own search engines and external engines with highly rich, chapter-representative searchable text of each book. Because it is UNCORRECTED material, please consider the following text as a useful but insufficient proxy for the authoritative book pages.

14 C H A P T E R 2 In the early 1970s, the peak of U.S. petroleum production and the ensuing “oil crisis” led to increased study of the energy efficiency of all transportation modes, including passenger rail. Even during this tumultuous period for the railroad industry, with several carriers in bankruptcy and the formation of Conrail and Amtrak, the rail mode was viewed as crucial to meeting future demand for energy-efficient transportation. Along this theme, in 1974 the U.S. DOT organized a conference on “The Role of the U.S. Railroads in Meeting the Nation’s Energy Requirements” (U. Wisc. 1974). The conference highlighted the opportunity presented by the energy efficiency of the rail mode. Research from this era concerned itself mainly with fuel economy and overall energy demand. Several new lightweight passenger trainsets were developed and tested during this period, and the Association of American Railroads began extensive research into the energy efficiency of trains. The resulting Train Energy Model, although developed primarily for freight applications, also allowed for the most detailed simulations of passenger trains to date. As scientists have reached consensus on the relationship between human activities and global warming, more recent research has focused on the potential for reducing greenhouse gas (GHG) emissions from the transportation sector through passenger rail. This research has examined technologies and alternative energy sources with the potential to directly reduce GHG emis- sions and improve energy efficiency. Recent research also has examined the role of passenger rail in reducing highway congestion, leading to improved efficiencies among all passenger-related surface transportation modes. This chapter summarizes past research into passenger rail energy efficiency. The chapter starts with a discussion of different efficiency metrics and commonly cited averages for various modes. This discussion is followed by summaries of more detailed studies of the efficiency of passenger rail, either alone or relative to competing modes. Finally, the chapter provides a summary of international benchmarks of conventional and high-speed passenger rail efficiency. 2.1 Passenger Transportation Energy Efficiency 2.1.1 Units of Measurement Several metrics are used to describe the efficiency of passenger transportation. Transportation productivity can be defined in terms of passenger-miles, seat-miles, train-miles or vehicle-miles. Accordingly, where diesel-electric propulsion is used, the efficiency of passenger rail systems can be described using passenger-miles per gallon, seat-miles per gallon, train-miles per gallon, and vehicle- miles per gallon. Passenger-miles per gallon—a metric used to describe the energy efficiency of individual pas- senger trips—is calculated in relation to system ridership and load factor (the percentage of seats occupied with passengers). Because the value of passenger-miles dominates calculations for this Passenger Rail Energy Efficiency Research and Benchmarks

Passenger Rail Energy Efficiency Research and Benchmarks 15 metric, differences in ridership can disguise variations in the inherent base efficiency of different transportation systems. Seat-miles per gallon measures the potential per-trip efficiency of a passenger transportation system. This metric is independent of ridership but is heavily influenced by the number of seats per vehicle (or railcar), however, and changes in seating configuration can overshadow the base efficiency of the system. Train-miles per gallon measures the overall energy efficiency of the entire train. This metric is independent of both ridership and the number of seats per railcar. It is directly correlated with the weight (and hence the length) of the train, however, with the result that longer, heavier trains appear to be less efficient as measured by this metric. Vehicle-miles per gallon describes the efficiency of each railcar (vehicle) in a train consist and is independent of both ridership and the number of seats. Although this metric is partially influ- enced by train length, with longer trains gaining efficiencies of scale and improved aerodynam- ics, vehicle-miles per gallon measures the inherent efficiency of the combination of rolling stock, infrastructure and operations present on a particular system. Which of these four metrics will be most appropriate for measuring and comparing energy efficiency across modes depends on the exact comparisons being made. For example, when comparing door-to-door trips on competing modes over specific routes, passenger-miles per gallon may provide the best comparison for a given ridership. When examining the potential of new technologies to improve efficiency, vehicle-miles per gallon may best describe the direct improvements to the inherent efficiency of the system. Per gallon metrics are familiar to the public given their association with automobiles and other light-duty passenger vehicles, but not all passenger travel modes obtain propulsion energy from internal combustion of gasoline and diesel fuel. For electric propulsion, similar metrics can be developed on a per kilowatt-hour (per kWh) basis. Equivalent metrics also can be developed per unit of natural gas or hydrogen fuel. When measured per kWh or per unit-mass or volume of fuel, however, the efficiency of systems that use different sources of energy can be difficult to compare. To facilitate these comparisons, efficiency can be expressed in terms of the unit of energy, such as per British thermal unit (Btu) or per kilojoule (kJ). The reciprocal of energy efficiency is energy intensity. Energy intensity quantifies the amount of energy required to provide one unit of transportation productivity. For passenger rail, energy intensity is usually described in Btu per passenger-mile (Btu/passenger-mi) or kJ per passenger- kilometer (kJ/passenger-km)—or their per seat equivalents—to provide reasonable working values without excessive use of decimals. A Note about Abbreviations NCRRP Report 3, NCRRP WOD 1 and the MMPASSIM spreadsheet tool use terms that, for convenience, may be variously abbreviated in text, tables, or figures, as follows: Term Abbreviations passenger-mile passenger-mi; pass-mi; pmi passenger-kilometer passenger-km; pass-km; pkm seat-mile seat-mi; smi seat-kilometer seat-km; skm train-mile train-mi; tmi train-kilometer train-km; tkm

16 Comparison of Passenger Rail Energy Consumption with Competing Modes 2.1.2 Average Passenger Transportation Energy Intensity The most widely available measures of passenger rail energy efficiency are those calculated on an annual gross average basis. This approach uses annual statistics, such as fuel or electric power consumed and transported passenger-miles, to estimate the energy efficiency and GHG emissions of passenger rail systems per passenger-mile. This method yields an effective high-level measure of passenger rail energy efficiency; however, annual gross average efficiencies should not be used to describe individual train runs and passenger trips. Each system, route, train run and passenger trip has unique characteristics that can cause the efficiency of that journey to significantly devi- ate from the annual average. To describe the performance of specific individual trips, the energy consumption and GHG emissions of each passenger mode on a specific route must be measured in the field or modeled via simulation. The latter approach has been taken for this project. The U.S. DOT Bureau of Transportation Statistics (BTS) reports the annual gross average pur- chased energy intensity of passenger transportation modes in the United States via the National Transportation Statistics database (see Table 2-1). Based on the information in the table, in 2011, on an annual gross average basis, passenger rail service operated by Amtrak was the least intense (most efficient) mode of passenger transportation. On average, light-duty vehicles (LDVs), such as automobiles and small trucks consumed nearly three times the energy of Amtrak trains per passenger-mile. (The most appropriate comparison to Amtrak service would be intercity buses; however, a measure of intercity bus energy use per passenger-mile isn’t available in the United States, as the BTS does not consider national passenger-mile statistics reliable.) A complicating factor is that the energy intensity figure for Amtrak includes both diesel- electric and electric motive power, whereas the competing modes are all powered by liquid fossil fuels. The presence of these two different sources of purchased energy—and the differing efficiency of their conversion to motion—clouds direct comparisons of purchased energy such as that made in Table 2-1. When viewed in isolation, electric locomotives can appear to be more efficient than diesel- electric locomotives. The energy consumed by an electric locomotive is readily measurable at the substation power meter or pantograph of the locomotive. Information about the fuel consumed in generating the electricity used by the locomotive is not readily available, however, and genera- tion fuel types vary by region. Thus, two commuter systems with identical ridership, equipment and operating characteristics also can have widely varying efficiency/emissions based solely on the source fuels used in electricity generation (e.g., coal in one region as opposed to renewable energy sources in a different region). Just considering the electric locomotive’s transmission efficiency, the work performed at the wheels is approximately 76–85% of that provided at the engine’s output shaft or pantograph (Lukaszewicz 2001). Mode Energy Intensity (Btu/pass-mi) Air 3,058 Light-duty Vehicle 4,689 Motorcycle 2,669 Transit Bus 3,343 Amtrak 1,628 Table 2-1. Energy intensity of passenger travel modes in 2011 (U.S. DOT BTS 2013).

Passenger Rail Energy Efficiency Research and Benchmarks 17 Diesel-electric locomotive efficiency also can be readily measured in terms of the fuel con- sumed by the locomotive. For a diesel-electric locomotive, this measure includes both the drive- train efficiency, which is common with the electric locomotive, and the combustion efficiency of the engine, which is not usually reported for the electric locomotive. The combined efficiency for a diesel locomotive is approximately 28–30% (Hoffrichter 2012). MMPASSIM overcomes the intrinsic difference in the efficiency of electric and diesel-electric locomotives by incorporating the energy input for fuels used to generate electricity for the region being simulated. Thus, while the Btu or kJ of energy reported by MMPASSIM for simulated diesel-electric or straight electric propulsion systems considers the fuel consumed in each case, the Btu or kJ reported for electrified systems will be higher than that reported by most reference literature or by any agency that measures the energy at the substation meter or on board the electric locomotive. As illustrated by the flow of energy through diesel-electric and electric locomotives (Figure 2-1), a tank or “meter-to-wheels” comparison ignores potentially significant losses associated with the generation and transmission of electricity from a remote generating site to the electric locomotive. The conversion of diesel fuel to energy for traction takes place on board the diesel-electric locomotive, so any losses that occur in conjunction with the conversion are incorporated into efficiency measurements. By contrast, losses associated with the generation and transmission of purchased electricity from a remote station to an electric locomotive occur before the electricity arrives at the train, so they generally are not reflected in measures of efficiency for the train. Measures of efficiency that are based on comparisons of the energy content of the purchased fuel to purchased kWh of electricity are thus skewed in favor of the electric train. In addition to differing energy conversion efficiencies, the energy and GHG emissions asso- ciated with transporting different fuels to the energy conversion site will vary by fuel type and by mode. These “upstream” energy/emissions differences are included in MMPASSIM via the Figure 2-1. Energy flow of (a) electric and (b) diesel-electric locomotives. (a) (b)

18 Comparison of Passenger Rail Energy Consumption with Competing Modes relationships shown in Argonne National Laboratory’s GREET (Greenhouse Gases, Regulated Emissions, and Energy Use in Transportation) model (Argonne National Laboratory 2012) with continental U.S. regional electricity generation fuel supply updated to the 2011 data from the U.S. Energy Information Administration. A complete well-to-wheels analysis on a per Btu basis does account for electric power gen- eration losses and differences in energy associated with fuel supply to provide a more accurate comparison between electric traction and internal combustion engines running on fossil fuels. Although electricity is provided by an interconnected grid of generation plants located in different states, the specific mix of generation sites supplying electric power to a commuter rail operator might differ from the regional groupings provided in the model. Thus, differ- ences in supply may occur that are different from those represented by the regions included in the model. For specific cases in which a user agrees to pay more for electricity from a new renewable energy site as a basis for getting GHG credits, the model includes a “wind-only” generation source that can be selected in place of the actual grid supply for the region from which the train would receive electricity. Solar and any other isolated electricity supply sources are not modeled, but as an open-source model, the user can add input data coefficients as desired. Although a well-to-wheels analysis accounts for geographic variation in electricity generation, it does not account for temporal variation within a region as different generating plants come on and offline during the course of a day. Because MMPASSIM is an open-source model, users can provide alternative regional supply characteristics and/or update the existing 2011-based supply grid. As outlined in the scope of work for NCRRP Project 02-01, this research does not consider lifecycle energy consumption and emissions from vehicle manufacture and disposal or infra- structure raw materials and construction (as may apply to different technologies or modes). 2.2 Previous Studies of Passenger Rail Energy Efficiency Many previous studies have quantified the energy intensity of passenger rail transportation based on analytical models, simulation and field data collection. To varying degrees, several of these studies have made direct comparisons between travel modes for specific routes. These studies also have taken differing approaches to considering ridership, access modes, time of day, trip purpose and upstream energy and emissions. The balance of this section provides summa- ries of several noteworthy studies, presented in chronological order. 2.2.1 FRA Study of Rail Fuel Consumption (Hopkins 1975) In May 1975, J. B. Hopkins of U.S. DOT completed a FRA study titled Railroads and the Environment—Estimation of Fuel Consumption in Rail Transportation. The first volume of the Hopkins study presented analytical models of fuel consumption for branchline freight service, line-haul freight service and passenger service. The freight analysis presented comparisons to highway mode, but passenger rail service was not compared in this manner. Hopkins developed a simple model of passenger rail fuel efficiency based on typical train resistance coefficients and locomotive fuel consumption for conventional passenger equip- ment of the era. Derived from first principles of power, tractive effort and train resistance, the model assumed a locomotive fuel consumption rate of 20.9 horsepower-hours per gallon

Passenger Rail Energy Efficiency Research and Benchmarks 19 to calculate efficiency in terms of seat-miles per gallon. The fuel efficiency calculation can be expressed as: S 8.16 10 vW ,mpg 7 s( ) = × where Smpg = passenger train fuel efficiency in seat-miles per gallon, v = train speed in mph, and Ws = train weight per seat in pounds. The simplified assumption of train resistance in pounds per ton embedded in this equation is assumed valid for speeds of 40 to 80 mph. The form of the equation indicates that fuel efficiency decreases as train weight per seat increases and also decreases as speed increases. The derived relationship reinforces the need for lightweight equipment to maintain efficiency as speed increases. Hopkins illustrated that, at 60 mph, conven- tional equipment operates at approximately 180 seat-miles per gallon while the more modern, lightweight equipment then being developed for high-speed rail (HSR) would operate at approxi- mately 550 seat-miles per gallon. Hopkins did not consider ridership or load factor to calculate the actual energy intensity of a passenger trip; however, Hopkins indicated that, because the weight of the passenger load is only 5% to 10% of train weight, the seat-miles per gallon metric is essentially independent of pas- senger load and can provide a better standard of measurement of the efficiency of the passenger train than passenger-miles per gallon. Hopkins acknowledged that, besides the two factors included in the equation, passenger train efficiency will vary with grade, train length, stop spacing and idle time. The report presented specific examples of the influence of each factor for trains with a given speed and weight per seat. Although the effect is small, increasing train length improves efficiency as the fixed drag and resistance of the locomotives is distributed over more railcars and seats in the longer train. Station stops and idling can greatly influence efficiency in practice; as much as 15% of passenger train fuel consumption occurs during idling time at stops. To reflect the HSR systems in service or being planned at the time of the study, Hopkins extended his model to trains designed to operate at speeds between 90 and 160 mph. This was accomplished by assuming that, at the design cruising speed for each train, the aerodynamic and rolling resistance of the train on a 0.5% grade exactly balances the tractive effort generated by the full rated horsepower of the trainset. The required horsepower-hours for the train to travel 1 mile are converted to gallons of fuel and then used to estimate seat-miles per gallon. Because the model is based on rated horsepower, the same approach is applied to diesel-electric, turbine and electric trainsets (Table 2-2). In the case of electric trains, this represents the equivalent amount of diesel fuel required to produce the required horsepower in a locomotive diesel prime mover. The values in Table 2-2 are for service on level grades and do not include stops or idling. Some designs were still in development in 1975, so the values presented in the table may differ from the final in-service condition. For example, the TGV prototype listed was powered by gas turbines at the time of the study, before later redevelopment as an electric train. An interesting comparison is between the U.S. TurboTrain and the longer Turbo version operated in Canada by Canadian National (CN). The CN train, with its greater seating capacity (326 seats versus 144 seats) and lower operating speed (95 mph versus 120 mph) is over twice as efficient in terms of seat-miles per gallon.

20 Comparison of Passenger Rail Energy Consumption with Competing Modes 2.2.2 U.S. DOT Study of Passenger Rail Energy Intensity (Mittal 1977) Completed in December 1977, a study with a scope similar to that of the 1975 FRA study by Hopkins was conducted by R. K. Mittal for the U.S. DOT and FRA. This study, titled Energy Intensity of Intercity Passenger Rail, examines the contemporary and future energy intensity of intercity passenger rail systems; the impact of new technologies and operating characteristics on the energy intensity; and the energy intensity of competing intercity travel modes. Mittal determined the energy intensity of passenger rail using statistical and analytical methods. The statistical approach used gross figures for annual fuel consumption and annual passenger- miles to calculate average Btu per passenger-mile for different train services. The analytical approach used known physical and engineering relationships of train performance to derive the energy intensity of particular trips via simulation. Fuel consumption rates of particular locomo- tives were used to calculate the energy consumption of the train service. The number of passengers on the train (expressed as the number of seats multiplied by the load factor) was used to deter- mine the energy intensity per passenger-mile. To compare electric and diesel-electric trains, Mittal calculated energy in Btu based on the consumed diesel fuel and the electricity input to the traction motors. The electricity input to the traction motors was based on an electric locomotive efficiency of 85%. Mittal provided some examples for which the energy intensity of the electric locomotives was derived from the energy consumed by the electrical generating station, based on an assumed generation efficiency of 35% and a transmission efficiency of 95%. Although this raised the energy intensity of the electric locomotives by a factor of three, the lower traction input values were used more extensively in the report. Mittal used analytical methods to calculate the energy intensity of different train consists at constant speed on level track. The values of various parameters were then changed to determine the sensitivity of energy intensity to factors such as train speed, passenger load factor, and train consist. Mittal then simulated the operations of trains over actual routes to determine the energy intensity of diesel-electric passenger trains between Albany, NY, and New York City, and of electric passenger trains between New York City and Washington, DC. Common diesel-electric locomotives of the time were included in the analysis: the General Motors Electro-Motive Division (EMD) E-8, SDP40F and F40PH, the General Electric (GE) P30CH and the Bombardier LRC. Mittal also considered the Ateliers de Construction du Nord Train (nation) Motive Power Cruise Speed (mph) Rated Power (hp) Seats Weight (tons) HP/Seat Estimated Seat-mi/ Gal. Metroliner (U.S.) Electric 110 5,900 246 360 23.9 65–95 TurboTrain (U.S.) Turbine 120 2,000 144 128 13.9 70–100 Turbo (Canada) Turbine 95 1,600 326 199 4.9 160–230 LRC (Canada) Diesel 118 5,800 288 452 20.1 115–170 Tokaido Shinkansen (Japan) Electric 130 11,900 987 820 11.4 180–270 HST (UK) Diesel 125 4,500 372 600 12.1 220–330 TGV001 (France) Turbine * 185 5,000 146 223 34.5 45–65 ER200 (U.S.S.R.) Electric 125 13,800 872 1,010 15.8 120–180 * The first TGV prototype used gas turbines before rising petroleum costs led to development of an all-electric design in 1974. Table 2-2. Estimated fuel efficiency of high-speed trains (Hopkins 1975).

Passenger Rail Energy Efficiency Research and Benchmarks 21 de la France’s (ANF) Turboliner, powered by a gas turbine; the Metroliner, with its electri- fied multiple-unit trainset; and three different electric locomotives: the GE E60CP, the Alstom CC14500 from France and the Swedish RC4a (later adapted to the AEM7 used by Amtrak). Although several of these locomotive types have been retired from service, several are still oper- ated in commuter and intercity passenger service today. The railcars considered in the study included refurbished 1950s-era passenger cars, newer single-level Amfleet coaches, and the lightweight coaches of the LRC and Turboliner trainsets. With the exception of the Turboliner, all of the railcars can still be found in service. Mittal also considered the presence of lounge, snack and meal-service cars in train consists in calculating energy intensity per seat and passenger-mile. The Mittal study offers many interesting conclusions. For a given load factor, passenger trains are found to reach their peak efficiency (lowest energy intensity) at cruising speeds in the range of 20 to 30 mph. These cruising speeds are lower than those at which light-duty passenger vehi- cles reach peak efficiency(50 to 60 mph). To provide intercity service times competitive with automobiles, passenger trains must operate outside their most efficient speed range. Mittal confirms the finding of Hopkins that the efficiency of passenger trains can be improved by increasing the number of passenger coaches in the train consist. This is found to be par- ticularly important for trains of conventional passenger cars hauled by heavier diesel-electric locomotives and less important for lightweight equipment such as the LRC. Lounge and snack cars are found to negatively impact passenger train efficiency, but Mittal suggests such passenger amenities are required to satisfy passenger demand and maintain an adequate load factor. At low load factors, passenger rail becomes very inefficient. Mittal verifies the assumption made by Hopkins that the weight of the passengers has little impact on passenger train fuel consumption. Energy intensity per train-mile or vehicle-mile is nearly the same under full-load and partial-load factors. Mittal makes an interesting comparison between the analytical passenger train energy inten- sity at a constant cruising speed of 65 mph and the energy intensity derived from simulating actual train operations on route grade profiles with speed restrictions and station stops. The real operating environment greatly increases the energy intensity of the passenger trains per seat- mile compared to steady cruising at 65 mph (Table 2-3). For the diesel-electric train consists, Propulsion Motive Power Cruising Energy Intensity (Btu/seat-mi) Cruising Speed (mph) Operating b Energy Intensity (Btu/seat-mi) Average Operating Speed (mph) Diesel-Electric E-8 443 65 820 49.3 P30CH 378 65 582 50.5 SDP-40F 412 65 555 50.5 LRC 289 65 528 50.4 Gas Turbine Turboliner 881 65 1,956 50.3 Electric a CC14500 365 65 963 68.3 Metroliner 310 65 1,019 78.4 a Electric energy intensity is based on input to traction motors and not on energy consumed at the power plant. b New York City–Albany, NY, for diesel-electric and gas turbine trains; New York City–Washington, DC, for electric trains. Table 2-3. Energy intensity of constant-speed cruising and actual service operating cycle (Mittal 1977).

22 Comparison of Passenger Rail Energy Consumption with Competing Modes the energy intensity increases by a factor of 50% to 100% under real operating conditions. The electric trains, with their higher operating speeds and more rapid acceleration, are approxi- mately 200% more energy intense under real operating conditions. This finding highlights the need for calculation of passenger train energy efficiency based on actual routes and specific train operations. Mittal’s data in Table 2-3 can be compared with the Hopkins data shown in Table 2-2. Hopkins estimates the efficiency of the LRC as 115 to 170 seat-miles per gallon and the effi- ciency of the Metroliner as 65 to 95 seat-miles per gallon. These numbers are equivalent to 670 to 991 Btu per seat-mile for the LRC and 1,200 to 1,750 Btu per seat-mile for the Metroliner. The values fall slightly above the in-service values presented by Mittal. However, given the assumptions for operating speed and grade made by Hopkins, they appear to offer reasonable agreement. The high energy intensity of the Turboliner compared to the diesel-electric and higher-speed electric trains parallels the low efficiency of the gas turbine TGV prototype in the Hopkins study. Mittal also surveyed the literature to determine the energy intensity of intercity passenger travel by aircraft, automobile and bus (Table 2-4). Historical trends of fuel consumption and load factor were considered for each mode. Comparison across modes suggested that passenger rail has the potential to operate with an energy intensity per seat-mile lower than that for auto- mobiles and aircraft. Based on the passenger rail load factors observed during the era, however, the energy intensity of passenger rail per passenger-mile was found to be greater than that of some competing modes. Bus was found to be the most efficient passenger transportation mode, followed closely by compact automobiles. At the time, passenger rail could not attract enough riders to take advantage of its potential efficiency. Mittal suggests the best way to improve the efficiency of passenger rail is to increase the load factor by attracting more ridership. Increased ridership can be accomplished by such things as (a) decreasing travel time (through improving track condition to allow higher operating speeds); (b) increasing the frequency of train departures; (c) improving the quality of service (e.g., seating density, amenities); and (d) lowering the cost of travel (i.e., ticket price). Mittal recognized the complex interactions between these factors and conducted additional analy- sis to determine the effects of making track improvements to decrease travel time. Although increased ridership and load factor resulting from reduced travel time worked to improve efficiency, an increase in operating speed had an overall mixed effect on energy intensity. Elimi- nation of speed restrictions and their associated acceleration events to create a more uniform speed profile tended to improve train energy efficiency, but increasing maximum operating speed increased aerodynamic drag and decreased efficiency. For the given routes and their Table 2-4. Energy intensity of intercity passenger transportation modes (Mittal 1977). Mode Possible Energy Intensity (Btu/seat-mi) Energy Intensity with Load Factor (Btu/pass-mi) Auto—Compact 1,100 1,900 Auto—Average 1,600 2,650 Bus 500 1,100 Air—Wide-body 3,000 5,500 Air—Current Fleet 3,600 6,500 Rail—Intercity 1,000 3,500 Rail—Metroliner 1,000 2,000

Passenger Rail Energy Efficiency Research and Benchmarks 23 distribution of speed restrictions, these two effects counteracted each other to maintain energy intensity for a given load factor. Thus, when the increased ridership from reduced running time was factored in, the track improvements and increased speeds decreased energy intensity per passenger-mile. Mittal acknowledged that this might not be the case for all corridors. If an increase in operating speed did not result in a large enough ridership increase, the efficiency of the passenger rail operation would ultimately decrease. It is important to note some differences between Mittal’s work and the scope of NCRRP Proj- ect 02-01. Although some of the same locomotives may still be in use, significant increases in energy efficiency have been obtained across all modes of passenger transportation over the last four decades. These changes make many of the results of the earlier study outdated. Also, Mittal does not consider the access and egress portions of the trip to make a true door-to-door com- parison. The provided per-mile comparisons may also neglect the influence of circuity where one mode takes a shorter, more direct route between two points than competing modes. Mittal’s values of Btu per passenger-mile (see Table 2-4) can be compared to the current national averages reported by the BTS (see Table 2-1). As would be expected through improved efficiency of modern equipment and increased load factor, both the air and rail modes currently exhibit much lower energy intensity per passenger-mile. LDVs, however, appear to be much less efficient than in Mittal’s estimates. This observation is surprising given the large advances that have been made in LDV fuel efficiency since 1977. It could reflect a combination of the BTS data considering congested city trips that are inherently less efficient than highway travel, and Mittal’s assumed occupancy rate of 2.4 persons per automobile. This occupancy rate is far greater than current FHWA statistics, which suggest that the current vehicle occupancy rate is closer to 1.15 persons per automobile. 2.2.3 Amtrak Northeast Corridor Testing, 1992–1993 (Lombardi 1994) As detailed by E. J. Lombardi (1994), during testing of new high-speed trainsets on the Amtrak Northeast Corridor in 1992 and 1993, energy consumption of different trains was measured and compared for trips between Washington, DC, and New York City (Table 2-5). A conventional train with one AEM7 and six Amfleet passenger coaches consumed 7,500 kWh for a single trip. The Swedish X2000 consumed 4,300 kWh of electricity, whereas the German ICE consumed 7,000 kWh. The two high-speed trainsets used regenerative braking to return approximately 17% of their consumed energy back to the grid. Lombardi attributed the higher energy con- sumption of the ICE train (compared to the X2000) to its higher horsepower, heavier weight and a faster operating speed, which exceeded the X2000 and AEM7 by 10 mph. The lower energy consumption of the two high-speed trainsets compared to the conventional train was attributed to aerodynamics. Train Energy Consumption a (kWh) Seats Possible Energy Intensity (Btu/seat-mi) Possible Energy Intensity (kWh/seat-km) AEM7 and Six Amfleet Coaches 7,500 504 226 0.041 Swedish X2000 (1L-5C trainset) b 4,300 355 183 0.033 German ICE (L-6C-L trainset) b 7,000 441 241 0.044 a Source: Lombardi 1994. b These are integrated trainsets (L = power car, C = coach). Table 2-5. Energy intensity of different trains on Amtrak’s Northeast Corridor.

24 Comparison of Passenger Rail Energy Consumption with Competing Modes The energy consumption data provided by Lombardi can be combined with trainset data to estimate the possible fully loaded (seat-mile) energy intensity of the different trainsets over the 225-mile route (Table 2-5). Although the trains all had very low energy intensity, the val- ues are based on the metered electric power consumption and do not consider the energy consumed in generating the electricity. When adjusted for load factor and source generation energy, these electric trainsets are more efficient than conventional trains hauled by diesel- electric locomotives. 2.2.4 Study of Metrolink Commuter Rail in Los Angeles (Barth et al. 1996) Barth et al. (1996) presented a paper titled Emissions Analysis of Southern California Metro- link Commuter Rail that estimated the emissions of a morning peak Metrolink commuter rail trip and compared them to the emissions of an equivalent automobile commute from Riverside to downtown Los Angeles, CA. Emissions for the line-haul portion of the commuter rail trip were determined by recording locomotive throttle settings for an actual train run and then multiplying by specific throttle-notch emissions factors developed during full-scale laboratory testing of the same locomotive model. The Metrolink study also included emissions from the station access segment of the com- muter rail trip. Surveys conducted on the train during a morning peak-period commuting trip asked passengers to detail their trip origin/destination, trip purpose, access mode, access dura- tion, access length, model of vehicle, egress mode, egress duration and egress length. The col- lected data were used to build a distribution of vehicle-trip profiles, and the emissions related to rail passenger access to and from Metrolink were calculated using the California Air Resources Board (CARB) EMFAC7F emissions model. A similar approach was used to determine the auto- mobile emissions of the highway trips required to transport the same number of commuters if they drove alone in their own vehicles instead of riding Metrolink. Total per passenger emissions for the Metrolink commute, including the train trip and access modes, were found to be less than those for the equivalent automobile commute when 300 highway commuters were compared to 300 train passengers that drove to the Metrolink station alone. Barth’s analysis was focused on local air quality/CAC emissions (i.e., carbon monoxide [CO], hydrocarbons [HC], nitrogen oxides [NOx], and particulate matter [PM] and did not assess GHG emissions). Compared to the automobile trip’s CAC emissions, the Metrolink commuter rail trip had lower CO and HC emissions, but higher NOx and PM emissions. The authors estimated that for the four morning trains under study, fewer than 100 rail passengers had to be diverted from the highway to result in a net reduction in CO and HC; 2,000 passen- gers for a net reduction in PM; and 1,500 to 2,000 passengers for a net reduction in NOx. Stated differently, each morning train produced CO and HC emissions equivalent to fewer than 25 automobiles, PM emissions equivalent to 500 automobiles and NOx emissions equivalent to 375 to 500 automobiles. Locomotive duty cycle data, as provided by Barth et al. in the paper, can be used to deduce the fuel consumption and energy intensity of the Metrolink commuter rail trip. (Locomotive duty cycle data also are given in Appendix A of NCRRP Report 3.) Using throttle-notch fuel consump- tion data for the EMD F59PH, each train run would consume 101 gallons of diesel fuel in direct propulsion and, given a minimum demand, another 25 gallons for head-end hotel power. Four morning trains carrying a total of 1,100 passengers would operate at 94 passenger-miles per gallon—equivalent to an energy intensity of 1,378 Btu per passenger-mile.

Passenger Rail Energy Efficiency Research and Benchmarks 25 2.2.5 Transport Canada Study of Passenger Transportation Emissions (Lake et al. 1999) Lake et al. (1999) conducted a study for Transport Canada titled Measures to Favour Passen- ger Modal Shift for GHG Reduction. This study characterized the GHG intensity of passenger travel modes for different origin-destination pairs in Canada. The analysis considered actual travel patterns, load factor and market share for different passenger travel modes. The per pas- senger estimates of emissions for most city pairs reflected the general expectation that the most favorable mode was bus, followed by rail, while auto and air were the least favorable modes. In the long-distance markets, however, emissions from rail were the highest or the second highest among all modes because of the need for sleeping and food-service cars on long-distance trains, which reduces the overall number of seats per railcar (and increases the train weight per seat). On certain routes where bus service had a relatively low load factor, automobile mode could produce the least emissions. Rail mode could also produce fewer emissions than bus mode on routes where strong rail ridership resulted in a high load factor. In addition to emissions data, the 1999 Transport Canada study included metrics for average energy efficiency of the different passenger transportation modes on the basis of passenger-miles and seat-miles (Table 2-6). On average, intercity bus was the most efficient mode. Comparison of rail intensity per seat-mile to bus intensity per passenger-mile indicates that, even if the pas- senger trains were operated at full capacity, they would only be more efficient than intercity bus at a typical load factor on a limited number of routes. 2.2.6 German Passenger Transportation Case Studies (Wacker and Schmid 2002) In their 2002 paper titled Environmental Effects of Various Modes of Passenger Transporta- tion: A Comprehensive Case-by-Case Study, Wacker and Schmid developed a complete energy consumption and emissions model for passenger transportation. Their methodology included the energy used in the main travel segment propulsion, access and egress mode propulsion, fuel production and supply, and in producing, maintaining and disposing of various transportation vehicles and infrastructure. Mode Energy Intensity (Btu/passenger-mi) Possible Energy Intensity (Btu/seat-mi) Intercity Bus 1,156 551 Rail—Average 2,114 -- Rail—VIA * Corridor East of Toronto -- 1,046 Rail—VIA Corridor West of Toronto -- 1,156 Rail—VIA Eastern Long-Distance Trains -- 1,542 Rail—VIA Western Long-Distance Trains -- 1,431 Air 3,665 -- Automobile 4,847 1,212 * VIA Rail Canada Table 2-6. Energy intensity of Canadian passenger travel modes in 1996 (Lake et al. 1999).

26 Comparison of Passenger Rail Energy Consumption with Competing Modes Besides considering access modes, the study by Wacker and Schmid is noteworthy for its consideration of trip purpose and time of day. The following trip scenarios were considered by appropriately adjusting the load factor of each transportation mode: • Commuter traffic: depart 8:00 a.m., return 4:30 p.m. • Shopping traffic: depart 10:00 a.m., return 12:00 p.m. • Leisure traffic: depart 7:30 p.m., return 10:30 p.m. • Sunday leisure traffic: depart 11:00 a.m., return 5:00 p.m. Case studies of a typical 20-mile interurban (IU) trip for leisure and commuter traffic were examined using different access/IU trip/egress mode combinations to reflect door-to-door travel. The modes included for one or more legs of the trips were automobile, transit bus, light rail transit, commuter rail, regional express trains, bicycle and walking. Direct energy use and emissions for automobile and bus transportation were calculated using the Handbook of Emission Factors for Road Transport (HBEFA). The Handbook database (a Microsoft® Access application) accounts for parameters such as traffic situation, road grade, motor system, and cubic engine capacity. For rail vehicles, the study used computer simula- tions to calculate energy consumption and emissions. Indirect energy consumption and emissions were calculated separately in three categories: (1) production of fuel and electricity; (2) vehicle production, maintenance and disposal; and (3) infrastructure construction, maintenance and disposal. The energy and emissions of vehicle production, maintenance and disposal were distributed evenly over the lifetime of the vehicle. For the case studies analyzed, light rail transit was the most environmentally friendly of the passenger travel modes. However, the results depended heavily on the time of day considered. For commuter traffic, trips involving rail were very competitive with those made by an automo- bile carrying four passengers. For mid-day leisure trips outside peak travel hours—when rail and public transit modes have a low load factor—the automobile with four passengers, or even with one passenger, was more efficient than rail and public transit. Additional long-distance intercity passenger rail case studies conducted as part of this research indicated that the German ICE high-speed train was more efficient and resulted in fewer emis- sions than the automobile with a single occupant. Only when the ICE train had a low load factor could an automobile with four occupants approach its level of efficiency. Wacker and Schmid also noted that for rail travel over longer distances, depending on the exact route and implemented technology, passenger rail could yield poorer results than an automobile with a single occupant. 2.2.7 Swedish Passenger Train Energy and Modal Comparison Study (Andersson and Lukaszewicz 2006) Andersson and Lukaszewicz (2006) led a Bombardier Transportation study to determine the average energy consumption and emissions of the modern passenger trainsets in Sweden. The study report, titled Energy Consumption and Related Air Pollution for Scandinavian Electric Passenger Trains, compared the measured energy consumption and related emissions for modern trainsets to older locomotive-hauled trains and averages for other modes of passenger transportation. For passenger trains, the energy calculations only accounted for energy used in propulsion, passenger comfort (i.e., head-end power, or HEP) and idling outside of scheduled service. The study did not include energy used in other activities, such as maintenance, operations of fixed installations or heating of facilities. Losses in the electric traction power supply system were accounted for by applying a scaling factor to the energy consumed by the electric trainsets at the pantograph. Regenerated energy from the train braking was subtracted from this total.

Passenger Rail Energy Efficiency Research and Benchmarks 27 However, losses at the power-generating station were not considered. This study considered the average emissions, marginal CO2 emissions, and average amount of electricity produced from renewable sources in estimating the emissions of the electric trains. Average energy consumption and emissions of other modes were not measured. Instead, averages were obtained from the Network for Transport and Environment, a Swedish non-profit organization aimed at establishing common base values for the environmental performance of transportation. Modal comparisons were made for trips from Stockholm to Gothenburg, Sweden, a distance of 283 miles (455 km) using the X2000 trainset (Table 2-7), and from Stockholm to West Aros, Sweden, a distance of 66 miles (107 km) using the Regina trainset (Table 2-8). For both routes, passenger rail exhibited the lowest energy intensity per seat-mile and per passenger-mile when load factor was considered. 2.2.8 Transport Canada Studies of Intercity Passenger Rail (2007 and 2010) In 2007, English, Moynihan and Lawson prepared a study for Transport Canada titled Assess- ment of Environmental Performance and Congestion Relief Benefits of Intercity Passenger Rail Ser- vices in Canada. The study, which was updated in 2010, used detailed spreadsheet models to simulate the emissions of rail, air, intercity bus and automobile trips. Confidential activity and load factor data were provided by all common carriers. Similar to the 1999 Transport Canada study, the work by English, Moynihan and Lawson determined that bus was the most efficient mode for all trips when load factor was used to calculate energy and emissions per passenger- mile (Table 2-9). Mode Possible Energy Intensity (Btu/seat-mi) Energy Intensity with Load Factor (Btu/pass-mi) CO2 Emissions (g/pkm) (g/pmi) NOX Emissions (g/pkm) (g/pmi) Rail (6-car X2000) 231 423 7 11 16 26 Air (Boeing 737-800) 1,757 2,800 130 210 600 968 Bus (Euro 3 Emissions) 373 1,098 53 85 360 581 Automobile (Mid-size car) 714 1,921 87 140 40 65 g/pkm = grams per passenger-kilometer; g/pmi = grams per passenger-mile Table 2-7. Energy intensity of passenger modes—Stockholm to Gothenburg, Sweden (Andersson and Lukaszewicz 2006). Mode Possible Energy Intensity (Btu/seat-mi) Energy Intensity with Load Factor (Btu/pass-mi) CO2 Emissions (g/pkm) (g/pmi) NOX Emissions (g/pkm) (g/pmi) Rail (3-car Regina) 165 478 8 13 18 29 Bus (Euro 3 Emissions) 412 1,208 59 95 409 660 Automobile (Mid-size car) 714 2,031 93 150 43 69 g/pkm = grams per passenger-kilometer; g/pmi = grams per passenger-mile Table 2-8. Energy intensity of passenger modes—Stockholm to West Aros, Sweden (Andersson and Lukaszewicz 2006).

28 Comparison of Passenger Rail Energy Consumption with Competing Modes The only route on which rail mode efficiency approached bus mode efficiency was for the service on Vancouver Island, where VIA Rail service was provided by a self-propelled diesel multiple-unit (DMU) railcar. For the longest transcontinental trip, Toronto to Vancouver, the improved load factor and distribution of the energy-intensive aircraft landing and take-off cycle over sufficient cruising distance brought the efficiency of the air trip within the range of the rail mode (with sleeping accommodations) and the auto mode. 2.2.9 Spanish Passenger Train Energy and Modal Comparison Study (Alvarez 2010) A. G. Alvarez conducted a study in 2010 titled Energy Consumption and Emissions of High- Speed Trains that compared the efficiency of conventional rail and HSR to competing modes of transportation on 10 different routes in Spain. The comparison was made via simulation software calibrated for rail operations in Spain. The analysis considered the actual distance traveled by each mode between a particular origin-destination pair because the shortest and longest modal paths could differ in length by as much as 30%. The comparison also used known load factors for each transportation mode and route in Spain to determine energy consumption and emissions per passenger-kilometer. Alvarez acknowledged the difficulty of comparing the electrified modes of transportation to other modes, and to each other, given that the emissions factors of power generation systems could vary between regions and also temporally within the same region. On seven of the 10 routes analyzed, the high-speed train produced fewer emissions than any other mode (Table 2-10; not all routes are shown). On the other three routes, the conventional Origin-Destination Distance (mi) Energy Intensity with Load Factor (Btu/passenger-mi) Rail Air Bus Auto Victoria–Courtenay * 140 1,596 28,737* 1,290 3,530 Ottawa–Montreal 116 2,518 10,727 860 3,530 Toronto–Montreal 335 1,699 4,308 880 3,530 Toronto–Vancouver 2,776 2,047 2,369 921 2,477 * The Victoria–Courtenay air service involved two short flight segments via Vancouver, as a direct service was not available. Table 2-9. Energy intensity of passenger modes—selected Canadian routes (English et al. 2007). Route and Shortest Distance Auto Bus Air Conven- tional Rail High- Speed Rail CO2 Emissions (g/passenger-mi) Madrid–Barcelona (302 mi) 209 48 235 57 46 Madrid–Alicante (223 mi) 197 54 263 46 53 Madrid–Valladolid (101 mi) 141 58 N/A 64 41 Average of 10 Routes 182 52 272 61 43 Energy Intensity with Load Factor (Btu/passenger-mi) Average of 10 Routes 2,635 659 2,965 1,427 1,043 Numbers are derived from original metric units. Table 2-10. Emissions and energy intensity of passenger modes— selected routes in Spain (Adapted from Alvarez 2010).

Passenger Rail Energy Efficiency Research and Benchmarks 29 train produced the lowest emissions because of the circuity in HSR routing between these points, which were more directly connected by the conventional rail network. On average, the conven- tional train produced over 40% more emissions than the high-speed train, and emissions for air and auto were four to five times that of the high-speed train. In terms of energy efficiency, however, the bus was the most efficient mode, followed by HSR as the second-most efficient mode, ahead of the conventional train. In concluding that the high-speed train was more energy efficient than the conventional train, Alvarez’s findings defy the “power law” convention by which higher speeds require greater energy consumption than lower speeds. Alvarez suggests that this arose from other factors that varied when the high-speed and conventional trains were compared between the same origin and destination. Besides attracting a higher load factor, the high-speed trains operated on routes that were typically shorter, with a more homogenous speed profile and with fewer stops and curves. The high-speed trainsets were also designed to have less weight per seat and better aerodynamic performance than the conventional trains. In Spain, the high-speed trains operated on a 25kV DC electrification system, while the conventional trains operated on a less- efficient 3kV DC system. Finally, the faster running time of the high-speed trains reduced the total cumulative power consumption of hotel and auxiliary power services between origin and destination, increasing energy efficiency. 2.2.10 Study of U.S. Corridor Transportation Energy (Sonnenberg 2010) In 2010, A. Sonnenberg at the Georgia Institute of Technology completed a study of Transportation Energy and Carbon Footprints for U.S. Corridors. Sonnenberg compared transportation-related carbon emissions by applying an emissions inventory framework to intercity travel modes including automobile, bus, air and passenger rail. Similar to Mittal, Sonnenberg utilized a corridor-based approach, rather than a door-to-door assessment, to apply the framework to three future HSR corridors in the United States: (1) San Francisco– Los Angeles–San Diego, CA; (2) Seattle–Portland–Eugene in the Pacific Northwest; and (3) Philadelphia–Harrisburg–Pittsburgh, PA. To compare the emissions of all modes, Sonnenberg employed a full lifecycle assessment, including the upstream and downstream emissions of transportation activity. For electric trains, the regional electric power generation profile was considered in the emissions analysis. Because the objective of the work was to determine the most effective strategies for reducing the overall transportation greenhouse gas (GHG) emissions in each corridor, Sonnenberg examined the relative effectiveness of changes to the fuel economy of different modes, introduction of alter- native fuels, development of new forms of transportation such as higher-speed and HSR, and changes to policy such as carbon taxes. With the goal of examining changes in overall corridor emissions, Sonnenberg did not directly compare energy intensity or emissions of competing modes for equivalent trips. However, Son- nenberg indicated that the introduction of 125-mph rail service did little to reduce corridor emissions and may actually have a negative impact. The trains may not be fast enough to obtain significant diversions from the air mode. If sufficient train frequency was implemented to obtain significant diversion from the highway mode, the extra train runs decreased the overall rail load factor, offsetting any efficiency and emissions gains per passenger. Also, for the corridor in Pennsylvania where the predominant electricity source is coal-fired power plants, the addition of electric trains had a much more negative impact than on other corridors where a greater propor- tion of clean and/or renewable source fuels is used to generate electricity. A 200-mph rail service was found to have a positive effect on corridor emissions, but only on the order of 0.5% to 1.5% reduction. Although HSR travel times were more competitive, the

30 Comparison of Passenger Rail Energy Consumption with Competing Modes rail service frequency required to divert trips from competing modes results in a low rail load factor. Also, because the express HSR service did not serve some smaller intermediate stations, conventional rail service was retained at lower load factors due to the loss of passengers to the HSR system. With a decreased load factor, the emissions from passengers at intermediate points bypassed by the HSR system actually increased, offsetting gains for passengers diverted to the new system. Sonnenberg came to the interesting conclusion that the best way to improve overall corridor efficiency and emissions was through automobile-based strategies. Autos had a large share of the main trip segment miles traveled, and they also represented a large share of the access and egress trips to the air, bus and rail modes. Thus, improvements to auto efficiency and emissions would actually improve the door-to-door performance of all modes. 2.2.11 North Carolina Regional Rail Study and Modal Comparison (Frey and Graver 2012) Frey and Graver investigated in-service fuel consumption and emissions rates for the North Carolina Department of Transportation (NC DOT) Rail Division on Amtrak regional intercity rail service between Raleigh and Charlotte, NC. The study, summarized in a 2012 report titled Measurement and Evaluation of Fuels and Technologies for Passenger Rail Service in North Carolina, used actual field measurements to determine the potential fuel and emissions savings for rail transportation compared to automobiles between cities along the rail corridor. The study also examined the implications of substituting B20 biodiesel as an alternative fuel in place of regular ultra-low sulfur diesel. A portable emissions measurement system (PEMS) was used to measure locomotive emis- sions during “over-the-rail” testing in service from Raleigh to Charlotte. Throttle position data collected during the tests were used to calculate fuel consumption and average energy intensity over the route (Table 2-11). The values for passenger rail included fuel and emissions associ- ated with both the locomotive prime mover and the diesel-generator set used for hotel power functions. On average, the HEP unit was responsible for roughly 8% of emissions and energy consumption. The rail results included a route average and a separate peak average for the most efficient segment of the route. Mode Energy Intensity with Load Factor (Btu/pass-mi) CO2 Emissions (g/pass-mi) Rail (Corridor Average) 3,125 246 Rail (Greensboro–Charlotte, NC) * 2,806 221 Automobile (1 occupant/vehicle) 4,993 384 Automobile (1.69 occupants/vehicle) 2,954 227 * The Greensboro–Charlotte segment was reported to have the lowest rail energy/emissions intensity for all combinations of the five locomotives and five origin-destination pairs monitored. However, the NCRRP Project 02-01 research team notes that—from the data in Table 5-6 on page 82 of the Frey and Graver report—the number presented in the text (221 g/passenger-mile) is for Locomotive No. 1755, which is 20% higher than the locomotive cited by Frey and Graver (Locomotive No. 1865, at 184 g/passenger-mile) and 10% higher than the fleet average number for that segment (201 g/passenger-mile). The automobile intensity numbers Frey and Graver report are based on the average of the five origin-destination pairs monitored/reported. Source: Frey and Graver (2012), pp. 84 (Rail) and 87 (Auto). Table 2-11. Energy intensity of passenger modes—Raleigh–Charlotte, NC (Frey and Graver 2012).

Passenger Rail Energy Efficiency Research and Benchmarks 31 Comparable highway trips were simulated with the EPA Motor Vehicle Emission Simulator (MOVES) software. The authors also presented two values for automobiles: one for vehicles with a single occupant and one with 1.69 persons per vehicle to match Department of Energy assumptions. Passenger rail was more efficient than the automobile if each traveler on the highway was in a separate vehicle. With greater highway-vehicle occupancy, the efficiency of passenger rail and highway vehicles became nearly equal, with the automobile more efficient on average but rail more efficient on certain segments (e.g., Greensboro–Charlotte) that had high ridership. With one occupant per highway vehicle, the CO2 emissions produced by 10 automobiles were equivalent to those produced by 15 passengers on the train, indicating that rail mode effec- tively reduced emissions. As in the earlier Metrolink study, a significant increase in rail ridership would be required to reduce other emissions factors, such as NOx. The study by Frey and Graver also found that, although B20 biodiesel reduced CO emissions by 12% compared to the ultra- low sulfur diesel fuel, NOx increased by 15%, HC by 6% and PM by 32%. This study also concluded that travel time is a significant factor in determining the emission factors on in-service trips. A delay of approximately 5% of the scheduled travel time increased the rail emissions rates for NOx, HC, CO2 and PM by roughly 16% per passenger-mile. The authors did not consider highway congestion and delay but acknowledged that examining these factors could improve the comparison in favor of passenger rail. 2.2.12 FRA Improved Passenger Equipment Evaluation Program (Bachman et al. 1978) This simulation-based assessment of HSR corridors in the United States was undertaken through the FRA’s Improved Passenger Equipment Evaluation Program (IPEEP) in the late 1970s (Bachman et al. 1978). Although the equipment characterized in the IPEEP studies is less relevant now, the corridors remain relevant. As discussed in the case study section (Chapter 4) of this report, the researchers for NCRRP Project 02-01 drew from the IPEEP reports for gradient and speed data for many of the railway corridor simulations in the MMPASSIM. 2.2.13 Summary of Previous Studies The studies examined for NCRRP Project 02-01 are all consistent in that none of them showed the automobile to be the most energy-efficient mode of transportation (Table 2-12). Of the six studies that included bus transportation (including the three North American studies), four identified bus as the most efficient mode of transportation. When compared to bus, rail was only the most efficient in Germany and Sweden, where the combination of lightweight trainsets requiring less energy and fast, frequent rail service that is well integrated with other modes to increase load factor results in very efficient passenger rail operations. The extremely low energy intensity of rail service in Sweden is attributable to a combination of efficient equipment, a high load factor and a regional generation mix in Scandinavia that favors hydroelectric power. The North American values of passenger rail energy intensity vary greatly, from a low of 1,378 Btu/passenger-mile for Metrolink in Southern California to a high of 2,806 Btu/passenger- mile for Amtrak regional intercity service in North Carolina. The main source of variation between these values is likely the load factor, with the morning Metrolink commuter train rep- resenting a nearly ideal case of peak ridership while the NC DOT service represents average pas- senger loads. During the peak holiday season, ridership on the NC DOT passenger rail service may reach higher load factors that produce lower energy intensities per passenger-mile. This variability reinforces the need for modal comparisons of energy efficiency to not only consider specific case study routes but to also consider trip purpose, temporal variation in factors and the access modes used in making a specific trip.

32 Comparison of Passenger Rail Energy Consumption with Competing Modes 2.3 Domestic and International Efficiency Benchmarks Previous sections in this chapter summarized research and formal experiments designed to investigate passenger rail energy and efficiency, either in isolation or in comparison to compet- ing travel modes. In addition to providing context for the research conducted in NCRRP Proj- ect 02-01, the results of these past studies can assist in validating energy and GHG emissions outputs from the MMPASSIM tool. Because several of the studies do not reflect current oper- ating conditions, additional benchmarks of the efficiency and emissions of existing passenger rail systems are needed to compare the outputs of the model to real-world results. Additional published efficiency and emissions data from commuter, conventional intercity and HSR sys- tems around the world are summarized in the balance of this section. 2.3.1 Commuter Rail Data on the energy consumption of individual commuter rail operations within the United States can be obtained from the National Transit Database (NTD). The NTD is the primary national database for FTA statistics on public transit. Transit systems that receive FTA grants are required to report various annual revenue, expense, ridership, operating and safety statistics for inclusion in the NTD. Although the majority of information in the NTD relates to bus, light rail and heavy rail transit systems, the NTD includes data on 26 different commuter rail systems that fall within the scope of this study. Specific NTD values of interest for this study are annual energy consumption in gallons of diesel, gallons of biodiesel or kWh of electricity; and transportation productivity in terms of passenger-miles, train-miles and vehicle-miles. The reported values of train-miles and vehicle-miles Table 2-12. Summary of previous passenger rail energy efficiency research. Study Country Most Energy-Efficient Mode Passenger Rail Energy Intensity with Load Factor (Btu/pass-mi) Considerations Auto Bus Rail Air Load Factor Source Generation Access Modes Time/Trip Purpose Mittal (1977) U.S. DOT/FRA United States 2,000 Barth et al. (1996) Metrolink United States N/I 1,378 N/A Lake et al. (1999) Transport Canada Canada 2,114 N/A Wacker and Schmid (2002) Germany N/I N/I Andersson and Lukaszewicz (2006) Sweden 423 English et al. (2007) Transport Canada Canada 1,596 N/A Alvarez (2010) Spain 1,043 Sonnenberg (2010) United States N/I N/I Frey and Graver (2012) NC DOT United States N/I N/I 2,806 N/A N/I = Not included in this study. N/A = Not applicable to this study.

Passenger Rail Energy Efficiency Research and Benchmarks 33 include deadhead and non-revenue movements. The consumed fuel and electricity were con- verted into common units of energy and used to determine the energy intensity of each system in Btu per passenger-mile, per train-mile and per vehicle-mile. The specific types of equipment being operated on each system were researched to estimate the Btu per seat-mile for each com- muter rail operation. Because electric power consumption is reported as purchased electricity, it does not include generation losses from the mix of sources in the region. Thus, the data pre- sented in this section reflect a tank or meter-to-wheels analysis. Although the goals of this study are to avoid gross average measurements of efficiency and to develop metrics for specific trips, the energy intensity values derived from the NTD provide a baseline for more detailed analysis. Also, because a unique energy intensity can be calculated for each of the 26 commuter rail operations, the calculated values indicate the wide varia- tion in efficiency resulting from differing system route, equipment, operating and ridership characteristics. The systems were categorized into four groups based on the type of propulsion in use: diesel- electric locomotive-hauled trains, self-propelled DMUs, electric and mixed or dual-mode systems. The latter category includes systems that utilize dual-mode locomotives that switch between elec- tric and diesel-electric propulsion during a trip, and systems that operate different lines that each use different forms of propulsion. Given that statistics are reported at the system level and by line, ridership and operating statistics for the electric and diesel-electric operations of each system are combined in the NTD. Thus, it is not possible to calculate separate electric and diesel-electric performance metrics for the systems operating separate lines of each type. The weighted-average intensity of the 14 systems exclusively using commuter trains hauled by diesel-electric locomotives is 2,242 Btu per passenger-mile and 574 Btu per seat-mile (Table 2-13). Significant variation occurs, with two systems operating below 1,600 Btu per State System Energy Intensity (Btu/pass-mi) Possible Energy Intensity (Btu/seat-mi) CA Altamont Commuter Express 1,575 451 CA North County Transit District—Coaster 2,662 516 CA Peninsula Corridor Joint Powers Board—Caltrain 1,905 585 CA Southern Calif. Regional Rail Authority—Metrolink 1,914 495 CT Connecticut DOT—Shore Line East 13,946 1,491 FL South Florida RTA—Tri-Rail 2,787 741 MA Massachusetts Bay Transportation Authority 2,153 509 MN Minneapolis Metro Transit—Northstar 2,722 618 NM Rio Metro Regional Transit District—Rail Runner 2,656 700 TN Nashville RTA—Music City Star 6,881 909 TX Trinity Railway Express 3,322 770 UT Utah Transit Authority—Front Runner 4,247 627 VA Virginia Railway Express 1,525 726 WA Central Puget Sound RTA—Sounder 2,306 658 Passenger-mile-weighted Average 2,242 574 Table 2-13. Energy intensity of diesel-electric locomotive-hauled commuter rail systems in 2011.

34 Comparison of Passenger Rail Energy Consumption with Competing Modes passenger-mile while several systems operate in excess of 3,000 Btu per passenger-mile. Less variation is seen in the energy intensity per seat-mile, as this metric is independent of load factor and provides a better measure of the actual efficiency of the equipment, infrastructure and operations. An interesting comparison can be made between the Altamont Commuter Express, with the lowest energy intensity (451 Btu per seat-mile), and the Virginia Railway Express, with an above-average energy intensity (726 Btu per seat-mile). This comparison suggests that the equipment, infrastructure and operations of the former system are inher- ently 40% more efficient than those of the latter system. Based on the NTD, however, the Virginia Railway Express operates with almost twice the load factor of the Altamont Com- muter Express, resulting in a nearly equal passenger trip efficiency of approximately 1,550 Btu per passenger-mile. Many of the systems with the highest energy intensity tend to be newer commuter rail opera- tions that began service during the past decade. Presumably this is because the newer systems are not yet fully integrated into regional transportation and development patterns and thus operate at a lower load factor. However, examination of the possible energy intensity per seat-mile suggests that load factor alone cannot explain this result. The newer systems tend to have higher energy intensity per seat-mile compared to the older, established, systems despite often having newer equipment with greater seating capacity. Further analysis of the NTD suggests that the newer services are operating primarily two- and three-railcar trains, whereas the older services with larger ridership are operating more efficient trains from four to seven railcars in length. The high energy intensity value per seat-mile for the Nashville system shows the compound- ing effects of short trains and less-efficient second-hand equipment. It is not known why the Connecticut DOT Shore Line East trains exhibit such poor performance per seat-mile (the poor passenger-mile performance can be attributed to an estimated load factor of 10%). A final observation can be made by examining the energy intensity per seat-mile of Minne- sota’s Northstar, New Mexico’s Rail Runner and Utah’s Front Runner. These three systems all use the same types of locomotives and similar-size trains of the same bi-level passenger railcars. Operating with essentially identical trains, the systems have an energy intensity per seat-mile of 618 Btu, 700 Btu and 627 Btu per seat-mile, respectively—all within 13% of each other. With ridership and equipment normalized, the remaining variation between the systems can be attrib- uted to differences in infrastructure and operating patterns. The weighted-average energy intensity of the five systems that use DMU railcars is 2,319 Btu per passenger-mile and 659 Btu per seat-mile (Table 2-14). State System Energy Intensity (Btu/pass-mi) Possible Energy Intensity (Btu/seat-mi) CA North County Transit District—Sprinter 2,011 563 NJ New Jersey Transit—River LINE 2,104 648 OR TriMet *—Westside Express 4,512 1,254 TX Capital Metro—Austin MetroRail 2,318 606 TX Denton County Transportation Authority—A-Train 6,612 857 Passenger-mile-weighted Average 2,319 659 Average of Systems Using European DMUs 2,097 617 * Tri-county Metropolitan Transportation District. Table 2-14. Energy intensity of DMU commuter rail systems in 2011.

Passenger Rail Energy Efficiency Research and Benchmarks 35 The three systems that use modern European-designed DMUs (Sprinter, River LINE and Austin MetroRail) exhibit fairly consistent energy intensity values of 563 Btu, 648 Btu and 606 Btu per seat-mile. These values differ by less than 8%, which suggests that any differences in infrastructure and operations between the systems have little influence on the base efficiency of the DMU. At the time of data collection, the Denton County A-Train was still using refurbished vintage rail-diesel cars (RDCs) at higher energy intensity per seat-mile while awaiting arrival and approval of their modern DMUs. The Westside Express uses a domestic DMU and trailer that is fully FRA compliant and weighs more than the European DMUs. Overall, the values for the DMU systems indicate that they are not substantially more efficient than the average system operating with diesel-electric locomotive-hauled trains. However, the DMU is not intended for markets that can support operations of very efficient trains of four to six passenger railcars. When the three systems operating European DMUs are compared to the three newer commuter systems operating shorter trains (Minnesota’s Northstar, New Mexico’s Rail Runner and Utah’s Front Runner), the DMU is inherently more efficient per seat-mile and also, when load factor is considered, per passenger-mile. Thus, where possible under FRA waiv- ers, when starting up a new commuter rail service, it may be more energy efficient to use more frequent DMU service to build ridership before implementing longer, locomotive-hauled trains with greater capacity. This result differs from the work of Messa (2006), who, from an emissions perspective, con- cluded that DMUs or trains of double-deck DMUs pulling trailers would always produce fewer emissions than locomotive-hauled trains and would be competitive with the emissions from electric power generation for electric trainsets. The data presented by Messa suggest that DMUs consume 450 Btu per seat-mile. Because no operator currently using DMUs has been able to achieve this level of efficiency, this value may be somewhat idealized and may not account for true operating conditions. The weighted-average intensity of the two systems using electric propulsion exclusively is 1,196 Btu per passenger-mile and 291 Btu per seat-mile (Table 2-15). Both systems exhibit very efficient operations in this meter-to-wheels analysis. If these values were adjusted upwards to account for regional generation efficiencies, they might increase by as much as three times, and the electrified systems would not be as efficient as several of the commuter systems that use DMUs and diesel-electric locomotives. The weighted-average intensity of the five systems that use a mix of electric and diesel-electric propulsion is 1,462 Btu per passenger-mile and 405 Btu per seat-mile (Table 2-16). These five systems are among the largest in the United States; they have very high ridership, and they operate some of the longest trains. Comparison to diesel-electric and DMU systems is clouded, however, by the nature of the purchased electricity. Considering the regional generation mix will increase the electric portion of the energy consumed by as much as three times. Doing this will have a varying impact on the efficiency of each system, depending on its proportion of electri- fied train-miles. State System Energy Intensity (Btu/pass-mi) Possible Energy Intensity (Btu/seat-mi) IN Northern Indiana Commuter Transit District—South Shore 819 211 PA Southeastern Pennsylvania Transit Authority 1,271 306 Passenger-mile-weighted Average 1,196 291 Table 2-15. Energy intensity of electrified commuter rail systems in 2011.

36 Comparison of Passenger Rail Energy Consumption with Competing Modes For all of the data expressed in Table 2-13 through Table 2-16, standing passengers during peak periods may allow the actual energy intensity in Btu per passenger-mile for a specific trip to be lower than the possible energy intensity expressed in the tables as Btu per seat-mile. 2.3.2 Conventional Regional and Long-Distance Intercity Rail Table 2-1 presented the national average energy intensity for Amtrak service as 1,628 Btu per passenger-mile. This value is comparable to the average value for commuter systems that, like Amtrak, use a mixture of electric and diesel-electric propulsion. Disaggregating this national average to obtain data on specific services and routes is difficult because Amtrak does not routinely directly track fuel consumption for specific train runs. Instead, Amtrak monitors fuel purchases and deliveries at their fueling points around the system, and a spreadsheet tool is used to allocate this fuel to different train services based on train size, operating speed and grade profile on the route. The situation is more complicated for the Northeast Corridor, where the electricity consumed by different Amtrak services is mixed within the pool of power supplied to different commuter rail services. Amtrak does not have meters on each locomotive that report power consumption data to a central database for management purposes. Collection of route-specific fuel consumption is complicated by the lack of reliable and accu- rate fuel gauges on locomotives. Regular locomotive fuel gauges often are unreliable, with University of Illinois students involved in Amtrak fuel studies reporting that locomotive fuel tank gauges would often show more fuel in the tank at the end of a run than at the start. Amtrak fuel studies on specific routes utilize a locomotive equipped with special fuel monitor- ing equipment. The NTD includes information for two state-supported Amtrak intercity corridors: the Downeaster from Boston, MA to Portland, ME, and the Pennsylvania DOT’s Pennsylvanian and Keystone Service, operating between Pittsburgh, Harrisburg, and Philadelphia, PA, and New York City. A test of biodiesel fuel on the Heartland Flyer between Fort Worth, TX, and Okla- homa City, OK, provided information on annual fuel consumption for that regional intercity route (Smith and Shurland 2013). The NCRRP Project 02-01 study team attempted to obtain fuel consumption data for other specific state-supported regional intercity services from public data. State rail plans and passenger rail budgets were consulted, but in most cases the fuel costs were lumped together with crew and other operating expenses. State System Energy Intensity (Btu/pass-mi) Possible Energy Intensity (Btu/seat-mi) IL Northeast Illinois Regional Commuter Rail Corporation—Metra 2,150 562 MD Maryland Transit Administration—MARC 2,031 618 NJ New Jersey Transit Corporation 1,670 380 NY MTA Metro—North Commuter Railroad Company 977 349 NY MTA Long Island Rail Road 1,261 339 Passenger-mile-weighted Average 1,462 405 Table 2-16. Energy intensity of mixed electric and diesel commuter rail systems in 2011.

Passenger Rail Energy Efficiency Research and Benchmarks 37 The three corridors for which data are available per passenger-mile exhibit lower efficiency than the Amtrak national average (Table 2-17). The Piedmont has higher energy intensity per seat-mile than the other corridors. This may be partially due to the short train consist on the Piedmont, which has only two revenue coaches to offset the resistance of the café-lounge car and the locomotive. Trains on the other two eastern corridors average five or six revenue coaches plus a café car and locomotive. The Heartland Flyer consists of three bi-level coaches, a locomo- tive and a non-powered control unit. Several international benchmarks of conventional passenger train performance in regional and long-distance intercity service are available. To provide a global context, these benchmarks and others obtained from the literature are compiled with the results showing significant varia- tion (Table 2-18). Oum and Yu (1994) explained international variation in passenger rail efficiency as being the result of various organizational, policy, societal, service and financial objectives of each national Corridor Energy Intensity (Btu/pass-mi) Possible Energy Intensity (Btu/seat-mi) Source Amtrak (national average) 1,628 -- BTS (2011) Northeast Corridor Regional Train (AEM7 and 6 coaches) -- 226 Lombardi (1994) NC DOT Piedmont 3,125 1,505 Frey and Graver (2012) Downeaster 2,510 921 NTD (2001) PennDOT Keystone and Pennsylvanian 2,703 859 NTD (2011) Heartland Flyer -- 574 Smith and Shurland (2013) Table 2-17. Energy intensity of selected Amtrak corridors. Table 2-18. International benchmarks of conventional passenger rail service. Country Service Propulsion Energy Intensity (Btu/pass-mi) Source Canada VIA Corridor Diesel 1,699 English et al. (2007) Spain Intercity Electric 1,427 Alvarez (2010) Sweden Conventional (8-car train) Electric 590 Andersson and Lukaszewicz (2006) Conventional (4-car train) Electric 650 OTU * Multiple-unit Electric 379 Norway Signatur Long Distance Electric 434 UK InterCity Diesel 1,081 Dincer and Elbir (2007) Turkey Intercity Diesel 1,560 Italy Intercity Electric 274 Federici et al. (2008) Denmark InterCity Diesel 790 Jorgensen and Sorenson (1997) Regional Diesel 1,370 * Öresund Train Unit

38 Comparison of Passenger Rail Energy Consumption with Competing Modes passenger rail network. The authors developed an index to measure the overall efficiency of passenger rail operations and determined that the leading nations at the time were Japan, the United Kingdom, Sweden and the Netherlands, followed closely by Denmark and Finland. The least efficient nations were Greece and Belgium. Andersson and Lukaszewicz benchmarked the energy intensity of conventional Swedish passenger trains at 590 Btu per passenger-mile for electrified locomotive-hauled eight-car trains and 650 Btu per passenger-mile for electrified locomotive-hauled four-car trains. The authors also presented the energy intensity of several higher-speed electric trainsets operat- ing in Norway (Signatur) and between Sweden and Denmark (Öresund train unit [OTU] multiple unit). Dincer and Elbir (2007) benchmarked the efficiency of passenger rail service in Turkey at 1,560 Btu per passenger-mile and in the UK at 1,081 Btu per passenger-mile. Federici et al. (2008) benchmarked the electrified intercity passenger rail service in Italy at an energy intensity of 274 Btu per passenger-mile. Jorgensen and Sorenson (1997) reported that diesel passenger trains in Denmark have energy intensities ranging from 790 Btu to 1,370 Btu per passenger-mile, depending on the load factor of the region and type of service. Compar- ing these values to the values given in Table 2-17, there is clearly room for improvement of domestic passenger train efficiency. 2.3.3 High-Speed Rail Currently no domestic operations meet the International Union of Railways (UIC) definition of true HSR of 250 kmh (155 mph) over extended distances on dedicated lines, and only the Amtrak Acela meets the definition of 200 kmh (124 mph) for specially upgraded lines. Because Amtrak does not track the energy consumption of specific trains on the Northeast Corridor, no data are available to provide a benchmark for domestic electric high-speed train energy consumption. Several researchers, including Levinson et al. (1997) and Chester and Horvath (2008, 2010), have examined the case of true HSR and its associated energy consumption and emissions. These studies take a full lifecycle approach, so their treatment of energy consumption is at a high level and is based on averages of various European HSR systems. The authors conclude, however, that California HSR will need to achieve, at a minimum, medium load factors in order to achieve better energy efficiency and emissions reductions than competing modes. Von Rozycki et al. (2003) conducted a unique assessment of the energy consumed by the high-speed ICE service, designed to operate at 250 kmh, between Hanover and Würzburg in Germany. The authors calculated that the ICE train consumed 22.5 kWh of electricity per train- kilometer for traction and that an additional 1.35 kWh per train-kilometer were consumed by train on-board functions and amenities (Table 2-19). The authors also determined an amount of “overhead energy”—1.20 kWh per train-mile—consumed in servicing, maintaining and making up the train. Finally, the authors considered access to the station by car, with an energy con- sumption of 0.945 liters of petrol per passenger. The authors also considered the ICE train load factor to determine the energy consumption per passenger-km and adjusted this value to reflect the generation efficiency of traction power supply. The access mode trip makes up about 18% of the energy consumed in the HSR passenger trip. As noted by Bosquet et al. (2013), special considerations are required to model the energy consumption of high-speed trains. Energy consumption increases rapidly with increasing speed as illustrated by SYSTRA (2011) simulation results for energy consumed at the wheels for two different trainsets (Table 2-20). At higher speeds, considerable increases in energy consumption are required for small improvements in travel time. To cut the base (200 kmh) travel time in half, energy consumption more than triples. However, as noted by Garcia (2010),

Passenger Rail Energy Efficiency Research and Benchmarks 39 reducing travel time decreases the total energy consumed by on-board train services and ame- nities. The two trainsets considered in Table 2-20 are the French TGV-Réseau, with 375 seats, and the 11-car AGV-11, with 460 seats. The AGV-11, with its distributed power configuration, is more efficient than the TGV-Réseau, with its concentrated locomotive power. When the number of seats per train is considered, the AGV-11 becomes even more energy efficient than the TGV-Réseau. With railway operation in Europe divided between rail infrastructure companies and train operating companies that pay for track access and traction energy, the energy consumption of individual trains is monitored much more closely than it is in North America. Thus, actual mea- sured energy consumption of high-speed trains over particular route segments is more prevalent in the literature. Jorgensen and Sorenson (1997) presented measured energy consumption of German ICE and French TGV train runs (Table 2-21). Although the consumed energy largely parallels the average speed, it is also influenced by the number of intermediate stops on the seg- ment, the grade profile and the exact model of train in use on the route. As part of the New Lines Program, Network Rail in the UK conducted a study in 2009 analyzing the relative environmental impacts of conventional rail and HSR. This study has supplied many efficiency and emissions benchmarks for existing HSR systems around the world (Table 2-22). Component Consumption Energy Intensity (kWh/train-km) (kWh/train-mi) (kWh/pkm) (kWh/pmi) (kJ/pkm) (Btu/pmi) Traction 22.5 36.2 0.0731 0.1176 755 1,152 On-board Services 1.35 2.2 0.0044 0.0071 33.5 51 Train Make-up 1.20 1.9 0.0039 0.0063 25.1 38 Passenger Automobile Access (L/pass) (gal./pass) (g/pkm) (gal./pmi) -- 0.945 0.25 4.2 0.0024 184 281 Total 998 1,522 pkm = passenger-kilometer; pmi = passenger-mile; U.S. units derived from the metric units Table 2-19. Energy consumption of German HSR Hanover–Würzburg ICE corridor (Von Rozycki et al. [2003]). Table 2-20. Simulated energy consumption of two high-speed trainsets at different speeds (SYSTRA [2011]). Speed (kmh) (mph) Time to Cover 100 km (min) TGV-Réseau Consumption * (kWh/tkm) (kWh/tmi) AGV-11 Consumption * (kWh/tkm) (kWh/tmi) 200 124 30 8.25 13.3 7.31 11.8 250 155 24 11.19 18.0 10.52 16.9 300 186 20 16.25 26.2 14.36 23.1 350 217 17 21.31 34.3 18.83 30.3 400 249 15 27.08 43.6 23.92 38.5 tkm = train-kilometers; tmi = train-miles * The authors indicate that “calculation estimates energy at the wheel, neglecting transmission and rolling stock losses, hotel power, and so forth; the assumed infrastructure is perfectly flat and straight, with no wind. Acceleration and braking are not taken into account.”

40 Comparison of Passenger Rail Energy Consumption with Competing Modes Janic (2003) benchmarked both German and French HSR energy efficiency at 0.044 kWh per seat-kilometer for the TGV and 0.058 kWh per seat-kilometer for the German ICE. Andersson and Lukaszewicz (2006) benchmarked the Swedish X2000 high-speed passenger trains with energy intensity of 0.042 kWh per seat-kilometer and 0.077 kWh per passenger-kilometer. The energy intensity of high-speed trains in Spain was quantified as 0.051 kWh per seat-kilometer by Jorgensen and Sorenson (1997) and 0.190 kWh per passenger-kilometer by Alvarez (2010). Le Maout (2012) reported on the efficiency of several routes in Japan and France in terms of kWh per passenger-kilometer. Service Segment Segment Length (km) (mi) Average Speed (kmh) (mph) Energy Consumption (kWh/tkm) (kWh/tmi) ICE Hamburg–Hanover 178 111 146 91 20.8 33.5 Hanover–Göttingen 100 62 190 118 26.0 41.9 Göttingen–Kassel–Wilhelmshöhe 45 28 138 86 32.4 52.1 Kassel–Wilhelmshöhe–Fulda 89 55 178 111 32.9 52.9 Fulda–Frankfurt 104 65 114 71 22.3 35.9 Frankfurt–Mannheim 78 48 114 71 23.1 37.1 Mannheim–Stuttgart 107 66 160 99 25.7 41.4 Stuttgart–Ulm 93 58 95 59 19.6 31.5 Ulm–Augsburg 86 53 129 80 19.5 31.3 München–Augsburg 61 38 118 73 26.4 42.4 Average 24.1 38.8 TGV Sud Est Paris–Lyon (2 stops) 427 265 214 133 17.4 27.9 Paris–Lyon (3 stops) 427 265 200 124 17.7 28.5 TGV Atlantique Paris–St Pierre des Corps 221 137 240 149 22.0 35.4 St Pierre des Corps–Bordeaux 348 216 144 89 13.2 21.2 TGV-Réseau Paris–Lille 226 140 229 142 18.8 30.2 TGV Duplex Paris–Lyon (2 stops) 427 265 270 168 17.7 28.4 Paris–Lyon (3 stops) 427 265 270 168 18.0 29.0 Paris–Lille 226 140 229 142 19.0 30.5 tkm = train-kilometers; tmi = train-miles Table 2-21. Measured energy consumption of different high-speed line segments (Jorgensen and Sorenson 1997).

Passenger Rail Energy Efficiency Research and Benchmarks 41 Table 2-22. International benchmarks of HSR passenger service. Country Train/Service Seats Speed (kmh) (mph) Energy Intensity (kWh/skm) (kWh/smi) Energy Intensity (kWh/pkm) (kWh/pmi) Source UK Class 91 InterCity 225 536 200 124 0.035 0.056 -- 1 UK Class 390 Pendolino 439 225 140 0.033 0.053 -- 1 UK Hitachi Super Express 649 200 124 0.028 0.045 -- 1 UK Class 373 Eurostar 750 300 186 0.041 0.066 -- 1 France TGV Sud Est 347 300 186 0.044 0.071 -- 2 France TGV-Réseau 377 300 186 0.039 0.063 -- 1 France TGV Duplex 545 300 186 0.037 0.060 -- 1 France AGV 650 300 186 0.033 0.053 -- 1 Germany ICE-1 743 250 155 0.058 0.093 0.073 0.117 2 Spain AVE Series 1 329 270 168 0.051 0.082 0.190 0.306 3 Spain AVE Series 103 Velaro 545 300 186 0.039 0.063 -- 1 Japan Tokaido 700 1,323 270 168 0.028 0.045 0.028 0.045 4 Japan Tokaido N700 1,323 300 186 -- -- 0.023 0.037 4 Japan Tohoku E2 815 275 171 -- -- 0.026 0.042 4 Japan Tohoku E5 731 320 199 -- -- 0.026 0.042 4 Sweden SJ2000 320 210 130 0.042 0.068 0.077 0.124 5 pkm = passenger-kilometers; pmi = passenger-miles; skm = seat-kilometers; smi = seat-miles Sources: Network Rail (2009); Janic (2003); Alvarez (2010); Le Maout (2012); Andersson and Lukaszewicz (2006)